Wikipedia isn't so much a learning tool, it's a summary page. If you want to learn complex topics in detail, you really need to find a better source.

Some Wikipedia pages are the exception to that, but it's rare.

So... a Neural Net?

I found Wikipedia very helpful back when I was at university in the "Theoretic Informatics 2" lecture, because believe it or not, my professor's lecture notes were much harder to follow even Unfortunately I forgot most of it by now.

A non-deterministic Turing machine is simply one where you have multiple possibilities to go from one state to another, in other words if you look at the state diagram instead of one arrow leaving each state it can be several. However just to understand what NP-complete means in general you do not need Turing machines at all. They only come important if you want to understand every little mathematical bit, as a mathematical model of computation itself, but unless you're a mathematician I don't see why you would want to.

(Also, for me one entire lecture was only on Turing Machines and then next year's only on complexity classes, so there simply is a lot of information to digest if you want to fully understand every little detail.)

I'm still fascinated by this though:

It's just that wikipedia has grown to be such an incorporated part of so many people's online internet experience. Google often points you right to a Wikipedia article. The thing is that an encyclopedia is supposed to be a place where anyone can go and gain knowledge without having to be an expert in the topic. The people who author the wiki know what they're talking about generally for the most part, but they do not have a clue how to go about properly explaining it to people.

I was reading about this last night, and had little clue what they were talking about :
http://en.wikipedia.org/wiki/Polynomial_hierarchy

Learning from an encyclopedia is the exception, not the norm. It's a reference, not a teaching material. It's a dictionary of concepts. It won't teach you how to use them.

I mean think about how all of your textbooks are structured. Prerequisite knowledge is either noted, or proceeds the chapter in the same book. Addition comes before multiplication. Multiplication comes before exponentiation.

Wikipedia stretches the line between encyclopedia and textbooks. Some of the articles are good enough to "teach" from. But that's not a requirement. There is no scholar deciding the scope of required material--every article stands on its own. You can go from multiplication, back to addition, and end up at Ph.D. level theories involving addition. There is no central "theme" or grouping to those topics to teach a specific theory or concept. There's also so much superfluous information that is useful to people in niche category X/Y/Z, but it tends to overwhelm and confuse someone who doesn't realize it's not important for learning the fundamental concept.

I find Wikipedia to be very good for general information, and very specific information on occasion, but not at all for learning fields. You'd end up with way too many holes in your understanding with no sense of whether you're grasping the depth of the material, or barely touching the surface.

It works well for "trivia stuff" like where Greece is, what Grapes of Wrath was about, or what a the chemical formula for a diesel fuel is. But other stuff--like learning the Central Limit Theorem--not so much. However, animated GIFs sure are amazing for showing many concepts.

Ah yeah, that's the part I had forgotten.

Okay, I think simple.wikipedia.org is a little too far in the other direction, now it's not advanced enough to learn very much. But it is useful for a simple direct overview of the subject.

I don't know but when I was a kid I used to love to read encylopedias and find out about things. And I don't remember it being all that difficult to understand a topic. Maybe their treatments were superficial, and that's why it was so easy, but then again maybe it was written specifically with ease of learning in mind.

I mean, if wikipedia isn't there to teach people new things, then what is it there for? Just as a reference for experts in the field? They could go a long way in making it more learner friendly, and using simpler language in introductions and more advanced topics later on in the articles.

Absolutely. I knew a ton of teachers who knew were very inept at breaking topics down into ways laymen could understand.

https://xkcd.com/214/

Hahaha lightweight!

This is what my computer looked like at the end of last night/3 AM.
{"name":"rOI13W6.png","src":"\/\/djungxnpq2nug.cloudfront.net\/image\/cache\/f\/a\/fac552b9dbfe107ba5fae2510c4c4381.png","w":1920,"h":73,"tn":"\/\/djungxnpq2nug.cloudfront.net\/image\/cache\/f\/a\/fac552b9dbfe107ba5fae2510c4c4381"}

Note the 163 = tabs open.

And I actually closed dozens before I went to bed, but I kept finding knew topics. Last night was a rare, really productive learning session.

Memory, optimizations, design-by-policy/contract, lots of D programming language, compilation of features missing in many languages (ala string slicing, digit grouping, etc), exceptions, stack-based allocation, data-driven programming, some DirectX/OGL mobile uglyness.

I'm much more likely to switch toward D. It's got way more helpful features than C++, yet lets you program closer to the metal than C++ and maybe C. It directly supports things like specifying structure alignment / padding. And they're apparently working toward making the entire D common library support "no garbage collection" mode. They have a rule about "if it looks like C++, it should function like C++" so there's no confusing duality of syntax meanings in your head. And I just found out the co-author is also the writer of the quintessential book "Modern C++: Generic Programming and Design Patterns applied" . There's lots more, but I don't want to fill this thread up.

WAIT, there's one insanely awesome one. You can write compile-time functions. No template insanity needed (though it supports templates). You can literally write imperative functions that write functions, or calculate numbers at compile time. You can also write in any function a "static if" statement that functions as a compile time decision without littering your code with preprocessor macros.

And default language support for Design by Contract which allows you to specify almost like an exception syntax the pre-conditions, post-conditions and invariants of your code. Previously, people made the same proposal for C++ and they rejected it--apparently helping us write clean, clear, safe code isn't in their budgets.

[edited]

I can't stop! They also support a built-in unit testing framework. You compile your program with a single flag, and it runs all unit-test{} blocks on your code. Everyone's all about unit tests these days but it's strange that nobody mentions that Design By Contract is the orthogonal complement to unit testing.

[edited again]

This is also possible in C++11 with constexpr.

I don't like D's struct vs class. It seems like a major design fault to have reference vs value types for a language with supposedly low level memory access possibilities.